A few weeks ago I started a new sensor and kicked off the BattleOfTheDexes I compared my xDrip Alogrithm against the Dexcom 505 Algorithm. (obviously its just one sample so nothing conclusive, but the results are in)
| METER | Dexcom | xDrip |
|---|---|---|
| 105 | 87 | 100 |
| 181 | 188 | 169 |
| 166 | 236 | 185 |
| 75 | 53 | 66 |
| 103 | 101 | 103 |
| 118 | 102 | 111 |
| 120 | 154 | 144 |
| 151 | 125 | 137 |
| 102 | 95 | 103 |
| 89 | 90 | 98 |
| 97 | 106 | 111 |
Overall MARDs:
- Dexcom: 15.5%
- xDrip: 8.7%
Note that Dexcoms actual MARD as calculated in studies is 9% I believe so its quite possible I had a bum sensor here. Also note that xDrip got one extra calibration value on the second day (because it asks for one if it thinks things are off at a time when it thinks the calibration will be the most helpful)
I stopped this first comparison here because once I restart the sensor on my Dexcom, xDrip winds up with a huge advantage as it already knows the slope from the last few days of calibrations wheras dexcom will be starting from scratch. I will do more comparisons each time I change sensors!
Want to know more about the project? Check it out HERE
All of the code is available and open source, if you want help getting it running just ask, Its constantly getting updated so either watch the repo in github or follow me on twitter @StephenIsTaken